Async write error

[Leif Halvarsson_2]

Hi,

I have got a problem with one of our rp2470 servers. The server is connected to a SAN and has two volumes located on a EVA 3000 array. I first noticed the problem when I tried to untar a large file on one of the SAN volumes. The process hung and was not possible to kill. I got the same problem when trying to untar the file on the other SAN volume but NOT on the internal disk.

In the syslog file I have a large numbers of errors like the following:

Mar 1 12:21:36 SRV00409 vmunix: SCSI: Async write error -- dev: b 3 0x040100, errno: 5, resid: 24576,
Mar 1 12:21:36 SRV00409 vmunix: blkno: 3200360, sectno: 6400720, offset: 3277168640, bcount: 24576.

I have 3 identical rp2470, connected to the same SAN and EVA array. On those servers I have not noticed any such problems.

OS release is HP-UX11i

Any ideas ?

[Hai Nguyen_1]

Leif,

Have you physically checked the hardware lights to quickly identify any equipment which may have some problem?

Hai

[A. Clay Stephenson · Accepted solution]

I assume that when you say three identical rp2470's that you also mean that they are patches to exactly the same level as well. Because you are seeing the same problem on two SAN volumes that tends to rule out a problem with a LUN/Volume. You might try increasing the IO Timeout but I'm doubtful of that. I am leaning towards a slightly faulty Fibre card or cable. If connected to a Fibre hub, you might try a different port. If you can arrange the downtime, you might move the Fibre card from the "Bad" box to a "Good" box and see if the problem moves with the card.

Errno 5 is so general that it is difficult to be very specific but it does point to some hardware or driver problem.

[Steven E. Protter]

fcmsutil /dev/td0 stat -s

change the dev to fit the actual situation.

If there are a lot of errors, fire up cstm, exercize the card. If it fails, plan and execute replacement of hardware.

SEP

[Leif Halvarsson_2]

Hi,
I had a spare rp2470 with a FC-HBA. I moved the internal disk to the spare server and changed HBA (WWN) in the host configuration for the EVA. And, as everything seems to be OK I think is was a hardware problem, most likley in the chain HBA-cable-switch port (GBIC).

Steven,
It was not possible for me to try your suggestions as I already had moved the disks but I will keep them in my mind.